Python-ideas April 2014

python-ideas@python.org

72 participants
41 discussions

Specify number of items to allocate for array.array() constructor

by Sven Rahmann

At the moment, the array module of the standard library allows to create arrays of different numeric types and to initialize them from an iterable (eg, another array). What's missing is the possiblity to specify the final size of the array (number of items), especially for large arrays. I'm thinking of suffix arrays (a text indexing data structure) for large texts, eg the human genome and its reverse complement (about 6 billion characters from the alphabet ACGT). The suffix array is a long int array of the same size (8 bytes per number, so it occupies about 48 GB memory). At the moment I am extending an array in chunks of several million items at a time at a time, which is slow and not elegant. The function below also initializes each item in the array to a given value (0 by default). Is there a reason why there the array.array constructor does not allow to simply specify the number of items that should be allocated? (I do not really care about the contents.) Would this be a worthwhile addition to / modification of the array module? My suggestions is to modify array generation in such a way that you could pass an iterator (as now) as second argument, but if you pass a single integer value, it should be treated as the number of items to allocate. Here is my current workaround (which is slow): def filled_array(typecode, n, value=0, bsize=(1<<22)): """returns a new array with given typecode (eg, "l" for long int, as in the array module) with n entries, initialized to the given value (default 0) """ a = array.array(typecode, [value]*bsize) x = array.array(typecode) r = n while r >= bsize: x.extend(a) r -= bsize x.extend([value]*r) return x

1 month, 2 weeks

Implicit string literal concatenation considered harmful?

by Guido van Rossum

I just spent a few minutes staring at a bug caused by a missing comma -- I got a mysterious argument count error because instead of foo('a', 'b') I had written foo('a' 'b'). This is a fairly common mistake, and IIRC at Google we even had a lint rule against this (there was also a Python dialect used for some specific purpose where this was explicitly forbidden). Now, with modern compiler technology, we can (and in fact do) evaluate compile-time string literal concatenation with the '+' operator, so there's really no reason to support 'a' 'b' any more. (The reason was always rather flimsy; I copied it from C but the reason why it's needed there doesn't really apply to Python, as it is mostly useful inside macros.) Would it be reasonable to start deprecating this and eventually remove it from the language? -- --Guido van Rossum (python.org/~guido)

2 years

Make __reduce__ to correspond to __getnewargs_ex__

by Neil Girdhar

Currently __reduce__<http://docs.python.org/3.4/library/pickle.html#object.__reduce__>returns up to five things: (1) self.__new__ (or a substitute) (2) the result of __getnewargs__<http://docs.python.org/3.4/library/pickle.html#object.__getnewargs__>, which returns a tuple of positional arguments for __new__<http://docs.python.org/3.4/reference/datamodel.html#object.__new__> , (3) the result of __getstate__<http://docs.python.org/3.4/library/pickle.html#object.__getstate__>, which returns an object to be passed to __setstate__<http://docs.python.org/3.4/library/pickle.html#object.__setstate__> (4) an iterator of values for appending to a sequence (5) an iterator of key-value pairs for setting on a string. Python 3.4 added the very useful (for me) __getnewargs_ex__<http://docs.python.org/3.4/library/pickle.html#object.__getnewargs_ex__>, which returns a pair: (1) a tuple of positional arguments for __new__<http://docs.python.org/3.4/reference/datamodel.html#object.__new__> (2) a dict of keyword arguments for __new__<http://docs.python.org/3.4/reference/datamodel.html#object.__new__> Therefore, I am proposing that __reduce__ return somehow these keyword arguments for __new__. Best, Neil

5 years, 10 months

OrderedDict literals

by Anthony Towns

Hi, I was re-reading some old threads about ordereddict literals to see if any of them had gotten anywhere. Amongst them, I came across a post by Tim Delaney: https://mail.python.org/pipermail/python-ideas/2011-January/009049.html that mentioned an odict literal of ['key': 'value', 'key2': 'value2'] could be confused with slice notation. >From a syntax point-of-view, that doesn't seem to be true (as mentioned in some of the replies to that thread), but it seems like you can abuse the similarity to make it a little easier to declare ordereddicts: from collections import OrderedDict class ODSlicer(object): def __getitem__(self, key): if type(key) is slice: key = [key] od = OrderedDict() for k in key: if type(k) is slice: od[k.start] = k.stop else: od[k] = k return od od = ODSlicer() print(od[1:2]) print(od["a":"b", "c":5]) print(od['a':'b', 'c':'d', ..., 'a':10, 'e':'f']) You could then replace: mydict = { 'foo': 'bar', 'baz': 'quux', } with: mydict = od[ 'foo': 'bar', 'baz': 'quux', ] if you need to convert a hardcoded dict into a hardcoded ordereddict. Works fine in python2.7 and python3.4. At this point, I'd like to note in my defence that this isn't called the python-good-ideas list :) What's the actual objection to supporting ['foo': 'bar'] odict literals? I saw Guido gave a -100, way back in the day, but no actual explanation for why it was distasteful? https://mail.python.org/pipermail/python-ideas/2009-June/004924.html Cheers, aj -- Anthony Towns <aj(a)erisian.com.au>

5 years, 11 months

A library for the deprecation of a function/class

by Stéphane Wirtel

With the CPython sprint, I was thinking about a lib to mark a function/class as deprecated. Example: #!/usr/bin/env python import sys import deprecation # from an external lib or from a stdlib module __version__ = (1,0) @deprecation.deprecated(python=(3,7,), msg='use foo_v2') def foo(): pass @deprecation.deprecated(program=(1,2), msg='use bar_v2') def bar(): pass @deprecation.deprecated(python=(3,7), msg='use inspect.signature()') @deprecation.deprecated(python=(3,7), to_use=inspect.signature) def getfullargspec(*args, **kwargs): pass The deprecated decorator should check the version of the software and the version of Python if asked with the arguments. it will raise warnings.warn with PendingDeprecationWarning or DeprecationWarning. Can be used in the documentation, via introspection. It's just an idea, there is no code, no specs but if you are interested I think I can try to propose a real solution with an external library and if this idea seems to be interesting, I will propose a PEP for 3.5 or 3.6. The interest of this lib, we can parse the source code and find the deprecated functions/classes with a small tool and the maintenance of the code should be improved. Please, could you give me your feedback? Thank you, Stephane -- Stéphane Wirtel - http://wirtel.be - @matrixise

5 years, 11 months

Re: [Python-ideas] Yet Another Switch-Case Syntax Proposal

by Lucas Malor

On 18 April 2014 16:58, Ed Kellett edk141-at-gmail.com | python-ideas-at-python.org| <baswps8ght(a)sneakemail.com> wrote: > case foo(): > > would have to become > > case (foo(),): > > to work as expected when foo() returned a tuple, which would mean > wrapping things in 1-tuples whenever you wanted to reliably match a > case that is determined dynamically. > To obviate this, instead of tuples case could check membership of CaseTuples. CaseTuple will be a class that extends tuple, identical to it, and a case expression list with commas will yield a CaseTuple. To use other iterables as case expression list you'll have to unpack them, or they will be matched for equality. On 18 April 2014 17:03, Joao S. O. Bueno jsbueno-at-python.org.br | python-ideas-at-python.org| <0ucjz7dbjt(a)sneakemail.com> wrote: > It may be just me, but I fail - in a complete manner - to see how this > syntax can offer any > improvement on current if/elif chains. It seems to differ from that > only by reducing > the explicitness, and the flexibility of the test condition that can > be used in any > of the subconditions > It's more simple to use when you can use it, as switch statement in the other languages. And it somewhat adheres to the DRY principle: why repeat the subject? If I'm checking what type of tarot card I have, why should I repeat every time I'm trying to identify a tarot card? I also thought about a syntax like this: "case" comparator case_expr ":" but IMHO it's too verbose for the typical uses of switch. If you want flexibility, you can always use if-elif. @Skip Montanaro: yes, switch statement is used in C also for code optimization. Frankly, I think this aspect is unimportant for CPython in the present time. On 18 April 2014 18:51, Andrew Barnert abarnert-at-yahoo.com | python-ideas-at-python.org| <3en9kh2cbt(a)sneakemail.com> wrote: > No it isn't. First, the "for ... in" keywords are not the same as just > "for" and a bunch of parens and semicolons. > My propose has elcase, that is not present in other languages. Also, notice that if you try to read the switch statement, or your Python > version, as English, it's nonsense. > Yes, the fact a case will behave differently for tuples and non-tuples will be difficult to translate in English. I think that with CaseTuple proposal it will be more understandable, since you have to explicitly unpack an iterable, or use elements separated by commas. Bash doesn't have separate "case" and "elcase" cases. After one case is > done, the rest are skipped, just as in most other languages. > But it has ;& and ;;& similarly to break and continue of C, that it's equivalent to case and elcase of my proposal. I don't see how skipping over any elcase but falling through to the next > case is in any way simpler than C. Well, because it's coherent with if-elif. See my last example in my first message. Well, then at least look at the limited form of pattern matching Python has > in the for and assignment statements and parameter matching, and maybe look > at how pattern matching is used with case statements in other languages; > don't try to suggest language designs based on guesses. > Excuse me? I know list comprehensions, lambdas and argument unpacking. And I do not think you can see what I do before I post a message. If so, you could see me googling before writing about something that I don't know very well or I don't remember very well. So don't guess about what I do or not do or know and not know, thank you. About pattern matching in the for statement, I really don't know what they are. .... and? Are you suggesting that if the switch expression is a string and > the case expression a compiled regex you could automatically call match > instead of testing for equality? If not, how is having regexp even relevant > here? And how are recursive functions relevant? > I'm suggesting to use if-elif with re module, if you want to use regular expression, and to use recursive functions if you want... recursive functions. To be more clear, IMHO switch-case is useful if it's simple. A generator expression is equal to anything except itself, and doesn't > contain anything. > You can convert it to an iterable. Probably an overkill, but you can do it. I don't know what you mean by "symbolic pattern" here. > For what I know (not too much), in Mathematica pattern matching can be used for symbols, and symbols can be used as identifiers: https://reference.wolfram.com/mathematica/guide/Patterns.html

5 years, 11 months

Yet Another Switch-Case Syntax Proposal

by Lucas Malor

Hello everybody. I firstly proposed this syntax in the python-list. As I already wrote there, I read PEP 3103 and I'm not completely satisfied by any of the proposed solutions. My first idea was a little different, this is my final proposal after a short but good brainstorm: switch_stmt ::= "switch" switch_expr "case" case_expr ":" suite ("case" | "elcase" case_expr ":" suite)* ["else" ":" suite] switch_expr ::= expression case_expr ::= expression_list - if case_expr is a tuple, the case suite will be executed if switch_expr is a member of the tuple - if case_expr is not a tuple, the case suite will be executed if switch_expr == case_expr - if a case_expr is checked, any subsequent elcase statements are skipped, and the next case statement is performed, of there's one. This is completely identical to if - elif. Example: briefing_days = ("Tue", "Thu") normal_days = ("Mon", "Wed", "Fri") switch day case normal_days + briefing_days: go_to_work = True day_type = "weekday" case normal_days: lunch_time = datetime.time(12) meeting_time = datetime.time(14) elcase briefing_days: lunch_time = datetime.time(11, 30) meeting_time = datetime.time(12, 30) else: go_to_work = False day_type = "festive" lunch_time = None meeting_time =None A simpler example: switch tarot case 0: card = "Fool" elcase 1: card = "Alan Moore" elcase 2: card = "High Priestess" <etc....> Some remarks: 1. switch is on the same line of the first case. This is because alternatives in PEP 3103 seems unpythonic to me 2. I decided to not use already existing keywords like "if" or "in", since this will be misleading 3. I preferred case / elcase instead of using break or continue keyword because: a. break / continue can confuse, since they are used in loop statements, and this is not a loop statement, as Ethan Furman pointed out in the other mailing list b. break / continue is useful only if it can be used not only as final command, so you can skip the rest of the current case suite. If it must be the final command, it's the same of case - elcase. IMHO there's no much real need to allow this c. case - elcase syntax is much more readable and less confusing of a break / continue d. case - elcase syntax is identical to if - elif, so people will be more familiar with it 4. notice that you can put a "case" statement after an "elcase" one. For example: switch cpu case "A6-6400K", "A8-6600K" clock = 3.9 elcase "A10-6790K", "A10-6800B" clock = 4.1 case "A6-6400K" tdp = 65 elcase "A8-6600K", "A10-6790K", "A10-6800B" tdp = 100 is equivalent to if cpu in ("A6-6400K", "A8-6600K") clock = 3.9 elif cpu in ("A10-6790K", "A10-6800B") clock = 4.1 if cpu == "A6-6400K" tdp = 65 elif cpu in ("A8-6600K", "A10-6790K", "A10-6800B") tdp = 100

5 years, 11 months

Python-specific Travis CI

by David Wilson

Hi there, In my travails to get py-lmdb (github.com/dw/py-lmdb) into a state I can depend on, I've been exploring continuous integration options since sometime mid last year, in the hopes of finding the simplest option that allows the package to be tested on all sensible versions of Python that I might bump into in a commercial environment. It seems this simple task currently yields one solution - Travis CI, except that by default Travis supports neither Windows, nor any version of Python older than about 3 months, without requiring serious manual configuration and maintenance. The library just about has a working Travis CI at this point – although it is incredibly ugly. Testing against all recent versions of Python (by "recent" I mean "anything I've seen and might reasonably expect to bump into while consulting, i.e. since 2.5") requires around 14 Debian packages to be installed, followed by installation of compatible versions of easy_install and Pip for each interesting Python version. The result just about works, but this week I'm again finding that it doesn't quite go far enough, as I begin working on a Windows binary distribution for my package. Windows is almost the same problem all over again - testing supported/sensible extension module builds requires 14 binary installations of Python (one for each of 32 bit and 64 bit), 3 distinct versions of Microsoft Visual Studio (that I'm aware of so far), and a compatible version of Windows to run it all on. So the work done to support CI on Travis more or less isn't enough. It seems the best option now is a custom Jenkins setup, and acquiring licenses for all the Microsoft software necessary to build the package. You may wonder why one might care about so many old Python versions in weird and exotic environments - to which I can't really give a better answer than that I've lost so much time working with packages that suffer from platform/version compatibility issues, that I'd like to avoid it in my own code. While it isn't so true in web development, there are pretty huge installations of Python around the world, many of which aren't quite near migrating to 2.6, never mind 3.4. This can happen for different reasons, from maintaining private forks of Python (perhaps in a post-audited state, where another audit simply isn't in the budget), through to having thousands of customers who have written automation scripts for their applications using an older dialect. In any case it's enough to note these installations do and will continue to exist, than to question why they exist.. So in the coming months I'll probably end up configuring a Jenkins, porting the Travis CI scripts over to it to a private Ubuntu VM configured as a Jenkins runner, and setting up a new Windows VM as a second runner, but it seems entirely wasteful that so much effort should be made for just one project. And that leads to thinking about a Python-specific CI service, that could support builds similar to Travis CI, except for all configurations of Python still in active use. One aspect of such a service that is particularly interesting, is the ability to centralize management of the required licenses – the PSF has a much greater reputation, and is much more likely to be granted a blanket license covering a CI service for all open source Python code, than each individual project is by applying individually. Licensing aside, having a 0-cost option for any Python project to be tested would improve the quality of our ecosystem overall, especially in the middle of the largest transition to date - the 3.x series. Following from the ecosystem improvement angle, there might be no reason a Python-specific CI service could not automatically download, build and py.test every new package entering PyPI, producing as side effect scoreboards similar to the Python 3 Wall of Superpowers (http://python3wos.appspot.com/). This would make a sound base for building future improvements for PyPI - for example a lint service, verifying "setup.py" is not writing to /etc/passwd, or uploaded eggs aren't installing a new SVCHOST.EXE to %SystemDir%. It could even make sense to have the hypothetical service produce signed, "golden" artefacts for upload to PyPI - binaries built from source in a trusted environment, free of any Bitcoin ransom-ware surreptitiously inserting itself into the bdist_egg ZIP file. There are many problems with providing a CI service, for example, the handling of third party commercial libraries required during the build process. An example of this might be producing a build of cx_Oracle, where Oracle's proprietary SDK is a prerequisite. Still, it would not require many resources, or even very much code, to provide a service that could minimally cover at least basic needs for purely open source software, and so the idea appeals to me. Any thoughts? David

5 years, 11 months

Design decisions about socket.sendfile()

by Giampaolo Rodola'

In http://bugs.python.org/issue17552 me and a bunch of other devs are currently discussing what kind of API socket.sendfile() should have and because it seems there are different opinions about it I though I should have brought this up here for further discussion. With issue 17552 I'm proposing to provide a wrapper on top of os.sendfile(), socket.send() and possibly TransmitFile on Windows. AFAIK the modules which could immediately benefit from this addition are ftplib ( http://bugs.python.org/issue13564) and httplib ( http://bugs.python.org/issue13559). This is the current function signature as of socket-sendfile5.patch: def sendfile(self, file, blocksize=262144, offset=0, use_fallback=True): """sendfile(file[, blocksize[, offset[, use_fallback]]]) -> sent Send a file until EOF is reached attempting to use high-performance os.sendfile() in which case *file* must be a regular file object opened in binary mode; if not and *use_fallback* is True send() will be used instead. File position is updated on return or also in case of error in which case file.tell() can be used to figure out the number of bytes which were transmitted. *blocksize* is the maximum number of bytes to transmit at one time, *offset* tells from where to start reading the file. The socket must be of SOCK_STREAM type. Non-blocking sockets are not supported. Return the total number of bytes which were transmitted. """ Debatable questions which were raised during the discussion on the bug tracker are: 1 - whether to provide a "use_fallback" argument which when False raises an exception in case os.sendfile() could not be used 2 - whether to provide a custom exception in order to signal the number of bytes transmitted as opposed to use file.tell() afterwards; for the record, I'm -1 on this. 3 - whether to commit the patch as-is without including Windows support via TransmitFile() and post-pone that for a later time in a separate issue 4 - (extra) whether to provide a "callback" argument which gets called against each block of data before it is being sent. This would be useful to apply data transformation or to implement a progress bar or something. Note if this gets added it would sort of conflict with the "use_fallback=False" parameter because in order to apply data transformation socket.send() must be used instead of os.sendfile(). So far #1 appears to be the most debatable question. -- Giampaolo - http://grodola.blogspot.com

5 years, 11 months

Re: [Python-ideas] 'from os.path import FILE, DIR' or internal structure of filenames

by Cameron Simpson

On 20Apr2014 14:41, anatoly techtonik <techtonik(a)gmail.com> wrote: >On Sun, Sep 29, 2013 at 2:26 AM, Cameron Simpson <cs(a)zip.com.au> wrote: >> -1 for any names commencing with __ (or even _). > >So you like it DIR, PATH As per-module/file predefined global "constants"? Possibly. Still -0, but open to argument on naming issues. >> -1 for new globals. > >So you want this: > from os.path import DIR, PATH I'm less good on these. That would import some staticish values (or are they to be functions?) which can't be right for the importing module. I suspect I'm misunderstanding your intent here. >> -1 because I can imagine wanting different nuances on the definitions >> above; in particular for DIR I can well imagine wanting bare >> dirname(abspath(FILE)) - semanticly different to your construction. >> There's lots of scope for bikeshedding here. > >Assuming that FILE is always absolute, how that: > dirname(abspath(FILE)) >is different from that: > abspath(dirname(FILE)) >? Depends what "absolute" means. When the source file was obtained by traversing a symlink, is FILE naive (ignored the symlink, just has to start at "/") or resolved (absolute path to FILE which does not traverse a symlink)? I can imagine use cases for both, and therefore the bikeshedding. >> -1 because this is trivial trival code. > >Code is trivial, but the problem is not. In particular, this code >doesn't work after os.chdir I'm fairly sure I acknowledged this. [...] Aha: On Mon, 30 Sep 2013 08:17:46 +1000 I wrote: Of course, chdir and passing paths to programs-which-are-not-my-children present scope for wanting abspath, but in the very common simple case: unnecessary and therefore undesirable. And I'm aware that modules-inside-zip-files don't work with this; let us ignore that; they won't work with abspath either:-) And what should happen for code from zipfiles? Something must happen, even if it is just NameError or the like, for "not supported". >> -1 because you can do all this with relative paths anyway, no need for abspath > >Non reliable because of os.chdir As mentioned above, yes. But very often irrelevant. (Hmm, maybe not in the general case of library code imported before a chdir and used afterwards.) Chdir's global nature (witin the process) is one reason casual use of it is discouraged; an amazing amount of stuff can be broken by a chdir. >> -1 because I can imagine being unable to compute abspath in certain >> circumstances ( certainly on older UNIX systems you could be >> inside a directory without sufficient privileges to walk back >> up the tree for getcwd and its equivalents ) > >Fear and uncertainty. Can you confirm that Python is able to launch script, but >abspath fails in these circumstances? There might be a bc break in 3.5 I'd need a suitable system today. On Linux, getcwd() has been a system call for a long time (I remember being slapped down on that very issue once). Let's try this Mac: $ mkdir -p fooo/baaa $ cd fooo/baaa $ chmod 0 .. $ /bin/pwd pwd: .: Permission denied $ So, yeah, not always reliable. This means that anything that presupplies names for [PHP-like __FILE__ and __DIR__] needs to compute it at need, not always. And the downside is that if these idoms, and importantly reliance on these idioms, becomes common then suddenly every Python program becomes a ticking time bomb for being run in an unusual-but-valid environment. Any kind of security conscious environment may require programs to run in constrained contexts. It is best to minimise the presumptions that a program makes in the name of convenience. More than once I've met devs whose code ran only in their (very very open/unsecured) dev environment and failed even in our staging environments. So I while it is rare, I don't think "/bin/pwd doesn't work" should be entirely ignored. Cheers, Cameron Simpson <cs(a)zip.com.au> Every particle continues in its state of rest or uniform motion in a straight line except insofar as it doesn't. - Sir Arther Eddington

5 years, 11 months

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

Python-ideas April 2014